nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Data evaluation

JohannesENS

Hello everybody,

I do have a data set with the following columns:

- Production Date/time
- Shift (3 per day)
- Location of interim storage (10 locations with 75 different options for each location)
- Batchnumber
- Product thickness (ranging from 100 to 1000)
- Product length (1 or 2)
- Quality inspection at the end of process for each product (Good or Bad)

The data of a row represents the data for one product.

At the end of the process I do receive good or bad products. What I want to do now, is to analyze the whole data set to find any similarities between the products for good or bad ones and find out what kind of issue leads to the bad products.

I already used the scatter plot matrix and tried to visualize the data in different ways by graph builder, but I want to go more into details.

What would you recommend to analyze this kind of data? Any suggestions or even tutorials which deal with this kind of problem?

Thank you for helping. If you need more information just let me know.

Best Regards,

Johannes

Accepted answers

Georg

I found the EDA in STIPS a quite useful training to learn these kind of analysis.

https://www.jmp.com/en_us/online-statistics-course/exploratory-data-analysis.html

Basically I would start with distribution platform, put all relevant variables in and look what happens when clicking on good/bad,

how the others are distributed.

You can continue with graphbuilder and drop variables in different dropzones, you can visualize much more than 2 variables ...

You can try fit y by x to see each variables influence as a single factor,

and can continue with fit model, putting in your y and all x, and see what happens.

At the end of course you could try to build a good model, to exactly understand the quantitative effect of each parameter.

But I think, the clue is to do it step by step, from the simple to the complex platform/model.

What also is great, to look at the sample files with example data and analysis of this kind.

See e.g. "Body Fat", it has plenty of analyses.

Open("$SAMPLE_DATA/Body Fat.jmp")

All comments

Mark_Bailey

There are many modeling techniques in JMP to explore relationships between the response and the potential factors. I like the Analyze > Predictive Modeling > Partition platform for such explorations. See this help page to get started.

Georg

I found the EDA in STIPS a quite useful training to learn these kind of analysis.

https://www.jmp.com/en_us/online-statistics-course/exploratory-data-analysis.html

Basically I would start with distribution platform, put all relevant variables in and look what happens when clicking on good/bad,

how the others are distributed.

You can continue with graphbuilder and drop variables in different dropzones, you can visualize much more than 2 variables ...

You can try fit y by x to see each variables influence as a single factor,

and can continue with fit model, putting in your y and all x, and see what happens.

At the end of course you could try to build a good model, to exactly understand the quantitative effect of each parameter.

But I think, the clue is to do it step by step, from the simple to the complex platform/model.

What also is great, to look at the sample files with example data and analysis of this kind.

See e.g. "Body Fat", it has plenty of analyses.

Open("$SAMPLE_DATA/Body Fat.jmp")

P_Bartell

For a book with many examples and case studies with a tie in with JMP as well I suggest purchasing "Visual Six Sigma...". Here's a link on SAS support books for purchase: Link to "Visual Six Sigma" . I think you'll find chapter 9 particularly similar to what you are attempting to do.

Byron_JMP

I like @Mark_Bailey 's answer, another similar method, which is kind of like cheating is to use the Analyze>Screening>Predictor Screening.

This tool will show you how each input variable contributes to the variation in the response variable. Its a quick check and is super useful.

JohannesENS

Thank you!

JohannesENS

Thank you